0
0
A groundbreaking study using AI-based analysis of nearly 10,000 pregnancies has identified previously unknown combinations of risk factors tied to serious pregnancy outcomes, including stillbirth. The findings could pave the way for more personalized and accurate pregnancy care, addressing significant challenges in predicting negative outcomes.
The study, led by Dr. Nathan Blue, MD, an assistant professor of obstetrics and gynecology at the Spencer Fox Eccles School of Medicine at the University of Utah, revealed that AI models are capable of detecting unexpected patterns in maternal and fetal characteristics. These patterns could significantly alter how risk is assessed and how care is personalized during pregnancy.
In this study, the researchers examined a dataset of 9,558 pregnancies, looking at a variety of factors such as blood pressure, social support, fetal weight, and medical history. The AI model identified new, previously unrecognized combinations of factors that contribute to complications such as stillbirth.
One of the most surprising findings was that, contrary to established knowledge, female fetuses with mothers who have pre-existing diabetes are at higher risk for complications compared to male fetuses. This discovery highlights how AI can reveal hidden correlations that even experienced clinicians might overlook.
The study also sheds light on the complexities of determining care for fetuses in the bottom 10% for weight, a group often considered borderline between healthy and at-risk. According to current clinical guidelines, intensive monitoring is advised for all fetuses in this group, which can be financially and emotionally taxing for families. However, the research found that the risk for adverse outcomes varies widely within this category. In some cases, the risk was nearly ten times greater than the average pregnancy, depending on factors such as fetal sex, pre-existing maternal diabetes, and fetal anomalies.
Dr. Blue notes that while the AI model uncovers correlations between variables, it doesn’t provide conclusive information on causality. He emphasizes that the AI tool does not replace clinical judgment but rather supports it, helping to ensure that decisions are based on reproducible and transparent data rather than subjective “gut checks” that can be influenced by human bias.
AI’s role in pregnancy care is especially crucial when it comes to rare scenarios. The researchers utilized an “explainable AI” model, which not only estimates risks but also provides a detailed breakdown of which factors contributed to those risk assessments. This transparency can help eliminate bias and improve clinical decision-making, making it more consistent and fair across diverse patient populations.
The next steps involve validating this AI model in broader, real-world populations to confirm its effectiveness in diverse clinical settings. Dr. Blue is optimistic that the tool will play a transformative role in personalizing care, allowing doctors to offer tailored recommendations based on each individual’s unique circumstances.
“This AI-based model can estimate risks specific to a person’s context in a transparent and reproducible way, something human judgment alone struggles to achieve,” said Dr. Blue. “Such tools could revolutionize how we approach pregnancy care, making it more individualized and precise.”
Disclaimer: While the study suggests promising findings, it is important to note that the AI model only identifies correlations and does not provide definitive causality. The tool should be used as an aid to clinical judgment, and further validation in diverse patient populations is necessary before widespread clinical implementation.
